nao

nao is presented as an open-source analytics agent (Apache 2.0 core, self-hostable, with GitHub source available). Data teams own a file-based context layer (database schemas, docs, business definitions, dbt semantics, rules and tests) that the agent reads, and the page states agent reliability grows with context. Users can ask questions in natural language, generate charts and monthly revenue views, build interactive live stories/dashboards, and get reports. The page also describes offline evaluation via nao test (scoring questions against expected SQL before context changes ship), online chat replay, and Context Recommendations that turn errors and downvotes into proposed fixes opened as pull requests. Enterprise options mentioned include bring-your-own LLM key with per-user and per-team budgets, self-hosting, SSO (Okta, Google, Microsoft Entra ID or any OIDC), row-level security and identity passthrough, and SOC 2 Type II. The page says it works with all warehouses, data stacks and LLMs, and it is backed by Y Combinator.

Key features

• Chat with data in Slack, Teams or Claude
• Build live dashboards, data apps and slides
• File-based context layer synced from warehouse, dbt and docs
• nao test offline evaluation of key questions against expected SQL
• Chat replay for online observability
• Context Recommendations proposing fixes as pull requests
• Bring your own LLM key with per-user and per-team budgets
• Self-hosting on your own cloud or nao Cloud
• SSO via Okta, Google, Microsoft Entra ID or any OIDC
• Row-level security and identity passthrough
• Open source core under Apache 2.0
• Works with all warehouses, data stacks and LLMs per the page

Best suited for

• Data teams that want to own and test analytics context
• Companies enabling non-technical staff to self-serve analytics
• Organizations needing self-hosted or permission-controlled BI
• Teams wanting an open-source, self-hostable analytics agent

Pricing

Not stated on the website